What Happens When You Use Dynamic AI Simulations Instead of Standard Videos for Learning? A 4th Grade Teacher’s Experience
The projector hummed in my 4th grade classroom last Monday, and I glanced up halfway through my go-to erosion video to see the same pattern I’ve seen a hundred times: three kids tracing dinosaurs in their science notebooks, Javi bouncing a pencil off his desk so hard the eraser chipped, and Mia passing a sticky note to her friend that says “this is boring”. I hit pause.
I’d gotten a flyer from our school tech coach the week before for a new tool that generates dynamic interactive simulations instead of pre-recorded standard videos. I’d tossed it in my desk drawer, assuming it was just another gimmicky edtech product that would add more work to my already full plate. But that day, with 25 minutes left in class and half the kids checked out, I pulled it up. I told the class we’d throw the lesson plan out the window for 15 minutes: instead of watching someone else’s experiment, they got to design their own.
Javi’s hand shot up immediately, which never happens. He said, “Last weekend I built a hill out of dirt in my backyard and put a bunch of grass and tiny pine seedlings on it. When I hosed it, barely any dirt came off. Why did this video’s hill fall apart with just three bushes?” We plugged his variables into the tool: 30% steeper grade than the video’s original hill, 80% vegetation coverage instead of 20%. It rendered the new simulation in 10 seconds, and we hit play. Rain fell at the same intensity as the original, and only a thin layer of dirt washed off the base. The whole class ooh-ed, which is not a reaction I get from most science videos.
Then Mia raised her hand. “There’s a new housing development at the top of the hill on my street. Every time it rains, mud covers our driveway. What if we add a paved parking lot at the top?” We did that, too. The simulation adjusted instantly, and when the rain hit, three times as much mud washed down as the original bare hill. By the time the bell rang, every kid could name two factors that impact how much erosion happens, something I’ve rarely gotten after a standard video lesson.
That afternoon, I picked up my 7-year-old son Leo, who’s been stuck on equivalent fractions for three weeks. I’d shown him three different standard YouTube videos that walk through cutting pizza into slices, but he kept getting stuck on a question none of the videos answered: “If you cut it into 12 pieces instead of 2, the pieces are smaller. How can half still be the same amount?” That night, I pulled up the same tool, pulled up a basic fraction simulation, and sat Leo down next to me at the kitchen table. I told him he got to pick how many slices we cut the pizza into.
He started with 2 slices, colored 1 in to mark 1/2. Then he dragged the slider to 12 slices, and the simulation split every existing slice right down the middle. The total colored area stayed exactly the same size, it was just split into 6 smaller pieces. He stared at the screen for a full minute, then leaned back and said, “Oh. The pieces are smaller, but there’s more of them. So the whole half is still the same.” That was it. No 20 minutes of me repeating the same explanation, no frustrated tears. He got it, from answering his own question.
Don’t get me wrong, it’s not all perfect. Last week I tried using it for a plate tectonics lesson, and the AI mixed up divergent and convergent boundaries when a kid asked what would happen if two ocean plates hit each other. I had to pause, correct it, and walk through the concept again before we kept going. It’s also easy to get off track: last week the class begged me to test what would happen if the erosion hill was made entirely of candy, and I caved, wasting 10 minutes of class on a silly detour that didn’t teach anything except that candy hills erode really fast.
Standard pre-recorded videos are still my go-to for most lessons. They’re reliable, they stick to the point, they don’t have random glitches, and I don’t have to worry about the AI getting a core concept wrong. But I’ve started changing how I use both, and there are three small, concrete changes I’ve made that work for me both in the classroom and at home with Leo. First, I don’t use simulations to introduce a new topic. I still start with a 5-minute standard video to lay out the basic rules, then use the simulation to answer all the random “what if” questions kids come up with. Second, I set a hard 15-minute time limit for simulations. Any longer, and we start to veer off into silly tangles that don’t move learning forward. Third, I always let the kids pick the variables, not me. If they want to test a silly idea, I let them do one silly one, then bring it back to the core question. It gives them buy-in, and they’re way more likely to remember the answer when they asked the question themselves.
Would I actually keep using it? Yeah, I would. But I wouldn’t throw out all my old standard videos to replace them. The biggest difference I’ve noticed is that standard videos answer the questions I, or the video creator, think are important. Simulations let kids answer the questions they actually care about. That’s a small shift, but it makes a big difference in who’s paying attention at the end of the day.
This week, Javi asked if we can test what happens if we add a retention pond at the bottom of our eroding hill. We’re doing that tomorrow. I don’t know if the simulation will get the science right, and I don’t know if we’ll finish our planned exit ticket before the bell rings. But I know every kid will be paying attention, and that’s more than I can say for most Monday mornings.
Please indicate: Thinking In Educating » What Happens When You Use Dynamic AI Simulations Instead of Standard Videos for Learning